The Agent Skills Directory

[PROMPT_INJECTION]: The skill architecture is susceptible to indirect prompt injection risks by design.
Ingestion points: The orchestrator processes user requests and project artifacts (such as DESIGN.md or blog content) which are then routed to specialized subagents like omd-kr-writer and omd-locale-adapter (see SKILL.md Stage 1-4).
Boundary markers: While the skill uses structured YAML envelopes for subagent communication, it lacks explicit safety instructions or delimiters to ensure subagents ignore instructions embedded within the processed artifacts.
Capability inventory: The orchestrator manages filesystem state by writing logs (.orchestrator.log) and review reports (.reviews/), and it triggers further agent actions based on subagent outputs.
Sanitization: There is no evidence of sanitization, filtering, or validation of content extracted from artifacts before interpolation into subagent prompts.

omd-orchestrator